Near singular matrix..
Posted: Sat May 03, 2014 1:27 am
When I try following rolling regression, eviews 8 gives me "Near singular matrix in "VAR VAR1A!IA!Y.LS(NOCONST) 0 0 DRM4 @ RMM4*P4A1 RMM4*M4A1" :(
Is there any method to avoid this error??
Thanks in advance m(_ _)m
Kohei
Code: Select all
for !i=1 to 413
smpl @first+!i-1 @first+98+!i
'産出価格と投入価格のOLS
equation b!i.ls value c soy
'残差の変数作成
b!i.makeresid rm!i
delete b!i
Series rmm!i=rm!i(-1)
Series drm!i=d(rm!i)
'残差をベクトルにして並び替える
stom(rm!i, vrm!i)
vector srm!i=@sort(vrm!i)
delete vrm!i rm!i
'並び変えたベクトルをシリーズに
mtos(srm!i, ssrm!i)
delete srm!i
'推計
for !y=1 to 70
for !j=16+!i-1 to 85+!i-1
genr p!ia!y=drm!i(-1)>=@elem(ssrm!i, !j)
genr m!ia!y=drm!i(-1)<@elem(ssrm!i,!j)
if !j=!y+!i+14 then
exitloop
endif
next
'ラグ次数0
var var1a!ia!y.ls(noconst) 0 0 drm!i @ rmm!i*p!ia!y rmm!i*m!ia!y
equation eq1a!ia!y.ls drm!i rmm!i*p!ia!y rmm!i*m!ia!y
'ssrのスカラー作成
scalar ssr1a!ia!y=var1a!ia!y.@ssr(1)
'q統計量
freeze(q1a!ia!y) var1a!ia!y.qstats(6)
scalar qq1a!ia!y=@val(q1a!ia!y(14,2))
delete q1a!ia!y
'係数スカラーの作成
vector vec1a!ia!y=@vec(var1a!ia!y.@coefmat)
scalar c1a!ia!y=vec1a!ia!y(1)
scalar d1a!ia!y=vec1a!ia!y(2)
delete vec1a!ia!y
'選別
If c1a!ia!y<1 and d1a!ia!y<1 and (1+c1a!ia!y)*(1+d1a!ia!y)<1 and qq1a!ia!y<10.645 then ssr1a!ia!y=ssr1a!ia!y else ssr1a!ia!y=100
endif
delete c1a!ia!y d1a!ia!y
'ssrの最小値の場所
vector(70) ssr1a!i
vector ssr1a!i(!y)=ssr1a!ia!y
delete ssr1a!ia!y
scalar iid1a!i=@imin(ssr1a!i)
next
next
Thanks in advance m(_ _)m
Kohei
